Docker-Provider

Version microsoft/oms:ciprod01072020 Version mcr.microsoft.com/azuremonitor/containerinsights/ciprod:ciprod01072020

Code change log

• Switch between 10255(old) and 10250(new) ports for cadvisor for older and newer versions of kubernetes

Customer Impact

• Node cpu, node memory, container cpu and container memory metrics were obtained earlier by querying kubelet readonly port(http://$NODE_IP:10255). Agent now supports getting these metrics from kubelet port(https://$NODE_IP:10250) as well. During the agent startup, it checks for connectivity to kubelet port(https://$NODE_IP:10250), and if it fails the metrics source is defaulted to readonly port(http://$NODE_IP:10255).

• Fix scheduler for all input plugins
• Fix liveness probe
• Reduce chunk sizes for all fluentD buffers to support larger clusters (nodes & pods)
• Chunk Kubernetes API calls (pods,nodes,events)
• Use HTTP.start instead of HTTP.new
• Merge KubePerf into KubePods & KubeNodes
• Merge KubeServices into KubePod
• Use stream based yajl for JSON parsing
• Health - Query only kube-system pods
• Health - Use keep_if instead of select
• Container log enrichment (turned OFF by default for ContainerName & ContainerImage)
• Application Insights Telemetry - Async
• Fix metricTime to be batch time for all metric input plugins
• Close socket connections properly for DockerAPIClient
• Fix top un handled exceptions in Kubernetes API Client and pod inventory
• Fix retries, wait between retries, chunk size, thread counts to be consistent for all FluentD workflows
• Back-off for containerlog enrichment K8S API calls
• Add new regions (3) for Azure Monitor Custom metrics
• Increase the cpu & memory limits for replica-set to support larger clusters (nodes & pods)
• Move to Ubuntu 18.04 LTS
• Support for Kubernetes 1.16
• Use ifconfig for detecting network connectivity issues

• Fix scheduler for all input plugins
• Fix liveness probe
• Reduce chunk sizes for all fluentD buffers to support larger clusters (nodes & pods)
• Chunk Kubernetes API calls (pods,nodes,events)
• Use HTTP.start instead of HTTP.new
• Merge KubePerf into KubePods & KubeNodes
• Merge KubeServices into KubePod
• Use stream based yajl for JSON parsing
• Health - Query only kube-system pods
• Health - Use keep_if instead of select
• Container log enrichment (turned OFF by default for TimeOfCommand, ContainerName & ContainerImage)
• Application Insights Telemetry - Async
• Fix metricTime to be batch time for all metric input plugins
• Close socket connections properly for DockerAPIClient
• Fix top un handled exceptions in Kubernetes API Client and pod inventory
• Fix retries, wait between retries, chunk size, thread counts to be consistent for all FluentD workflows
• Back-off for containerlog enrichment K8S API calls
• Add new regions (3) for Azure Monitor Custom metrics
• Increase the cpu & memory limits for replica-set to support larger clusters (nodes & pods)
• Move to Ubuntu 18.04 LTS
• Support for Kubernetes 1.16
• Use ifconfig for detecting network connectivity issues

Hotfix that fixes the agent crash seen in some 5.* azure kernel versions

1. Update prometheus config scraping capability to restrict collecting metrics from pods in specific namespaces.
2. Feature to send custom configuration/prometheus scrape errors to KubeMonAgentEvents table in customer's workspace.
3. Bug fix for init containers.
4. Bug fix for empty array being a valid setting in custom config in configmap.
5. Restrict kubelet_docker_operations and kubelet_docker_operations_errors to create_containers, remove_containers and pull_image operations.
6. Fix top exceptions in telemetry

1. Cluster Health Private Preview based on config map setting
2. Fix for promethus config map telemetry.
3. Telemetry for controller kind.
4. Update url to use one of the whitelisted urls for cp monitor telemetry.
5. Update custom metrics supported regions.
6. Configmap with clusterid for Application Insights
7. Update resource requests for replicaset to 110m and 250Mi.

• Prometheus custom metric collection using config map allowing omsagent to
  • Scrape metrics from user defined urls
  • Scrape kubernetes pods with prometheus annotations
  • Scrape metrics from kubernetes services
• Exception fixes in daemonset and replicaset
• Container Inventory plugin changes to get image id from the repo digest and populate repository for image with
  only image digest
• Remove telegraf errors from being sent to ApplicationInsights and instead log it to stderr to provide visibility for
  customers
• Bug fixes for region names with spaces being processed incorrectly while sending mdm metrics
• Add log size in telemetry
• Remove buffer chunk size and buffer max size from fluentbit configuration

04/23/2019 - Version microsoft/oms:ciprod04232019
This is a release for Azure AKS monitoring. This is not intended to be used for oms-solution.

• Windows node monitoring (metrics & inventory)
• Telegraf integration (Telegraf metrics to LogAnalytics)
• Node Disk usage metrics (used, free, used%) as InsightsMetrics
• Resource stamping for all types (inventory, metrics (perf), metrics (InsightsMetrics), logs) [Applicable only for AKS clusters]
• Upped daemonset memory request (not limit) from 150Mi to 225 Mi
• Added liveness probe for fluentbit
• Fix for MDM filter plugin when kubeapi returns non-200 response

Core dump fix for ContainerImageInventory provider

This is a release for Azure AKS monitoring. This is not intended to be used for oms-solution.

• Fix for closing response.Body in outoms
• Update Mem_Buf_Limit to 5m for fluentbit
• Tail only files that were modified since 5 minutes
• Remove some unwanted logs that are chatty in outoms
• Fix for MDM disablement for AKS-Engine
• Fix for Pod count metric (same as container count) in MDM